Qualitative and Quantitative Statistics
There are two different kinds of statistics.
One, and that which you are probably most commonly used to seeing is called enumerative (quantitative) statistics. Here, large samples are used and explicit questions asked - questions that people can answer with a yes or no, or questions that people can answer in a logical, rational, or rationalized fashion.
The reason that large samples are necessary using quantitative statistics is twofold:
1. Since people tend to talk with the same words about any product regardless of the brand, and further, to defend that which they have already said or done, it's necessary to use lots of people to get statistically significant differences in response.
2. Quantitative statistics develop percentage information and the samples have to be large enough so that this information can be projected to the entire universe - so you can know how many people have bought Brand A or Brand B and how many will say because of "price" or "store reputation," etc.
This kind of statistics calls for probability sampling - where everyone has an equal chance to be selected with everyone else - or some form of stratified sampling where all relevant segments are taken into consideration and sufficiently covered (in numbers) to develop projectable numbers for each segment and for the population as a whole.
Enumerative statistics are also sometimes called parametric statistics since they describe the parameters of the universe and obviously, the bigger your sample (or population), the more of the universe you include.
The second general kind of statistics is called qualitative or analytical statistics. Since instead of counting the numbers of people who do or say anything, you are dealing with attitudes and making inferences or conclusions based on those attitudes - finding out why, not how much - the sample is not based on probability methods at all.
In analytic statistics, you have to start from the basis of selecting relevant populations and you need use only enough people so that you can read differences in attitudes among the various segments of interest.
Of course, one of the things that is true when research is done this way is that you must be in contact with the world as it exists. This includes picking (or setting a quota for) the right population for the product area you are exploring - less by buying habits than by involvement. The other thing that is true, and absolutely basic to this kind of procedure, is that the test items, when used, must be administered at random among the selected population in order to generate accurate data.
For this reason, using analytic statistics, we put great emphasis on experiment design and questionnaire development and, making the assumption that most of our clients are relatively skilled at knowing about other people or they wouldn't be in business, offer data which should be as easily read by them as it is by us - not in terms of the statistics maybe, but in terms of the sense of the words or phrases that people have used to define the various evaluated objects.
If, in the end, you understand how different people feel about something and why they feel that way, you can often go on to predict how many will buy it because you can track the numbers of those who are responsive to your message, whether in demographic or other terms
Further, you can tell how to increase your buying public by pulling in those people who are "on the edge" of your market, and you can define them, too, in terms of demographics and attitudes; and say how many customers they will represent.
While all this is, of course, oversimplified, there are two basic points to remember in creating marketing, merchandising or advertising strategies.
1. When you ask "how many" - that is, using only quantitative statistics - you have to guess why, and often the why doesn't really show up.
2. When you ask "why" and understand the different "whys" - using analytic statistics - you can predict success or failure by tracking existing segments on existing products and then seeing how those same segments respond to your approach.
